@it
F O R E S T SERVICE
U S D E P A R T M E N T O F AGRICULTURE
P O BOX 245, BERKELEY, CALIFORNIA 94707
Forest and Range
Experiment Station
FINE FUEL MOISTURE M E A S U R E D A N D ESTIMATED
in d e a d Andropogon virginicus i n Hawaii
Francis M. Fujioka
Fujioka, Francis M.
1976. Fine fuel moisture measured and estimated
in dead Andropogon virginicus in Hawaii.
USDA Forest Serv. Res. Note PSW-317, 4 p.,
illus. Pacific Southwest Forest and Range Exp.
Stn., Berkeley, Calif.
Fuel moisture estimates generated by the National
Fire-Danger Rating System procedure were compared
with actual fuel moisture measurements determined
from laboratory analysis. Meteorological data required
for the NFDRS procedure were collected at two
heights to assess the effect of temperature and humidity lapse rates. Standard measurements gave the best
results, but use of a xeric hygrometer showed promise
of giving reliable fuel moisture estimates by a simple
linear regression.
Oxford: 431.2:812.211-015.
Refrieval Terms: fuel moisture estimation; National
Fire Danger Rating System; hygrometrix; xeric
hygrometers; sampling.
The ignition potential of fine fuels-that is, grasses
and other herbaceous plants-is critically dependent
on fuel moisture. The National Fire-Danger Rating
,System1 (NFDRS) provides a method for estimating
fine fuel moisture (FFM) in dead vegetation using
meteorological data. The r e l e ~ n tweather variables
are dry-bulb temperature, relative humidity, and
state-of-weather, here considered to mean cloudiness.
C
In the NFDRS computer algorithm,2 FFM is directly
proportional to the equilibrium moisture c on tent,^
which can be computed from the weather variables.
Presently, it is not known how well FFM estimates
approximate fuel moisture of dead vegetation in
Hawaii.
This study compared moisture content calculated
by the NFDRS method with moisture content of fuel
samples as determined in the laboratory. Temperature
and relative humidity were measured at two different
heights, to see what effect the actual vertical variation
of these variables had on fuel moisture calculations. A
new instrument, the ~ ~ ~ r o m e txeric
r i x hygrometer,
~
which employs a cellulosic element that responds to
changes in atmospheric humidity, was tested for its
ability to estimate FFM. Using a combination of
techniques, I generated four estimates for each observation time. Only one of these measurements conforms to NFDRS standard procedure, but I have
included the correlations between all estimates and
actual fuel moisture. The standard procedure yielded
the best estimate, but the results showed potential for
using the Hygrometrix hygrometer readings in a
simple linear regression.
METHODS
Paired observations were made of weather and fuel
moisture of broomsedge (Arzdropogon virginicus L.),
a major wildfire fuel in Hawaii. The sampling site
overlooked Tripler Military Reservation, on the lee
side of the Koolau Range. Mean annual rainfall in this
area is about 125 cm (50 inches). The broomsedge
grew on a southwest-facing slope, about 300 m (1000
ft) above mean sea level. A portable fan psychrometer
and a Hygrometrix xeric hygrometer were placed in
weather shelters set at two heights, one 0.25 m (10
inches) above ground, and the other 1.22 m (48
inches), the standard observing height.
At hourly intervals during daylight, the observer
collected three samples of dead fuel in airtight roundbottom 500 ml boiling flasks; he read the dry- and
wet-bulb temperatures from the psychrometer, and
noted the Hygrometrix index at each shelter. He also
estimated amount of cloud cover and recorded the
appropriate state-of-weather code.
The fuel samples were processed by xylene distillation to determine percent moisture content by dry
weight, and the average of the three values for each
collection time was talcen as actual fuel moisture
(AFM).
For the computed fuel moisture (CFM), two
estimates were made for each height, using two
different values for humidity: CFM, used the value
obtained from the fan psychrometer and CFM2 used
the value observed on the xeric hygrometer scale.
Only the CFMl estimate from the 1.22 m height
legitimately follows the NFDRS procedure. A total of
58 sets of weather measurements, 29 at each shelter
height, were used with 29 corresponding actual fuel
moisture measurements.
In the analysis phase, I computed the simple linear
correlation coefficients between each of the CFM's at
each height, and the AFM, and between the direct
xeric hygrometer readings and AFM. I also tested for
statistical differences between AFM and CFMl .
RESULTS AND DISCUSSION
The computed fuel moistures generally underestimated actual moisture content. Actual fuel
moistures, obtained by the xylene process, gave a
pooled standard deviation of 1.02 percent. Examination of the differences between AFM and each of the
fuel moisture estimates showed that CFMl at 1.22 m
was apparently best (table 1). It had the lowest
absolute mean difference, the lowest root-meansquare difference, and a comparably low standard
deviation of difference. This might have been expected, because the data used for the other estimates
were not obtained by the legitimate procedure. The
next best set of estimates was the CFM2 set at 0.25m.
Table 1-Statistical analysis o f differences between actual
fuel moisture and computer fuel moisture1 (AFM-CFM)
Shelter height
and CFM
estimates
Mean
difference
Root-meanStandard
deviation
square
of difference difference
0.25 m
CFMl
CFM2
-1.6
-1.2
1.1
1 .O
1.9
1.6
-0.7
-1.7
1.1
1.1
1.3
2.0
1.22 m
CFM 1
CFM2
'Humidity estimate for CFMl obtained from fan psychrometer; for CFM2from xeric hygrometer scale.
Table 2 gives the correlation coefficients computed between each of the CFM's and AFM, and
between the direct xeric hygrometer readings and
AFM. All the correlations were significant at the 1
percent level. The highest correlations were associated
with use of direct xeric hygrometer readings, but the
correlation coefficients were not vastly different from
each other. A chi-square test of the l~omogeneityof
the coefficients in table 2, using Hotelling's z* transforms, showed no significant difference at the 10
percent level.
Table 2-Correlation coefficients of values for actual
fuel moisture and computed fuel moisture
Shelter
height
CFM 1
CFM2
Hygrometris
index
The high correlation between AFM and the direct
xeric hygrometer reading suggests that a linear regression model may reliably estimate AFM from xeric
hygrometer observations Cfig. I ) . The coefficient of
determination, r 2 , is 0.87.
The plots of fuel moisture over time Cfig.2) show,
as does table I, that the CFM's generally underestimate AFM. Results for CFMl at 1.22 m, which
was calculated following the NFDRS procedure, are
of greatest interest here. The effect of the "sunnycloudy" variable on CFM apparently complicates the
results. The procedure directs the observer t o report
cloudy conditions outside the period from 1000 to
1500 LST, even though sunny skies are actually
observed. Otherwise the fuel moisture estimates are
unduly depressed (the NFDRS algoritlun decreases
relative humidity by one-fourth of its reported value,
l:Fi
2 August 1972
'
'P
:12
+.
LO
,.."...\
.......
;:>.-r.,L.-.le
\;
9
Xeric index
........."
.
,/
l I00
;.
,;,
,.-.
.-
;
;
\.'... ;
;
\.'{,
\./
7
0700
.
1500 0700
1100
1500
1900 0600
1230
1900
Figure I - A simple linear regression of actual fuel
moisture content (percent of dry weight) on xeric
index gives the equation as shown.
when sunny conditions are observed). This rule was
initially overlooked, and tlie calculations produced
CFM-AFM differences that were twice as large as they
should have been. To see how good CFMl would be
if we always assumed cloudy conditions, I recomputed fuel rnoistures for all clear sky data points,
assuming cloudy conditions for each, and obtained
the differences given in figure 3.
Sunny slties were observed 11 times in the
sampling period. In 7 of these instances, the CFM's
matched closer to the AFM when cloudy conditions
were assumed. Significantly, the 4 times when the
sunny slty estimates were better than the assumed
cloudy estimates were those when AFM was low
(<lo%). The mean difference between sunny CFMl
and AFM was - 1.9; assuming cloudy conditions, the
mean difference between CFM, and AFM was reduced in magnitude to 0.6. Tlie first difference is
significant at the 5 percent level, the second is not
(10 d.f.).
Tlie data suggest that the information on cloudiness does not improve the fuel moisture estimate
computed from the algorithm. On the other hand,
when sunny slties are ignored, CFM tends to exceed
AFM. How significant is the overestimate? At the 5
percent level, the difference between mean CFMl
(assuming cloudy slties) and mean AFM was not large
enougli to reject t l ~ ehypothesis that mean CFM,
<mean AFM. However, the probability of a Type I1
error (acceptance of a false hypothesis) in this partic-
-I-:
0600
1230
T ~ m eo f
.............
C FM,
day (HST)
- .- .- .-
CFM,
- AFM
Figure 2-Fuel moisture estimates (CFM's) were
compared with actual fuel moisture (AFM) on 3
days. Estimates from lower shelter (0.25 m) are
shown at left, from standard shelter height
(1.22 m) at right. Local apparent noon (sun at
relative zenith) occurs at about 12:35 this time of
year. Fuel moisture estimates were obtained from
fan psychrometer (CFMl) and from xeric hygrometer scale (CFM*).
ular test is 0.78. More sampling is needed t o reduce
this margin of error, and to investigate the seasonal
variability (if any) of differences between computed
and actual fuel moistures.
There is much appeal in obtaining fuel moisture
estimates quickly, using the xeric hygrometer readings in a simple linear regression. However, a further
study must test the effectiveness of tlie regression.
Continuous sampling should also be considered, to
investigate local diurnal variations in fuel moisture.
3
o
2-
O
0
-
0
0
0
I
NOTES
Sunny (observed)
Cloudy (assumed)
x
-
-
0
0
0
1
6x
8
lb
0
1'1
I2
13
1 AFM
14
X
Figure 3-The difference between fuel moisture
estimates and actual fuel moistures, assuming
cloudy conditions for actually sunny conditions,
was determined, with results as shown.
' Deeming, John E., James W. Lancaster, and Michael A.
Fosberg. 1972. National Fire-Danger Rating System. USDA
Forest Serv. Res. Paper RM-84, 165 p., illus. Rocky
Mountain Forest and Range Exp. Stn., Fort Collins, Colo.
Furman, R. William, and Robert S. Helfman. 1973. A
computer program for processing historic fire weather data
for the National Fire-Danger Rating System. USDA Forest
Serv. Res. Note RM-234, 12 p., illus. Rocky Mountain Forest
and Range Exp. Stn., Fort Collins, Colo.
Fosberg, Michael A., and John E. Deeming. 1971. Derivation o f the I - and 10-hour timelag fuel moisture calculations
for fire-danger rating. USDA Forest Serv. Res. Note RM-207,
8 p. Rocky Mountain Forest and Range Exp. Stn., Fort
Collins, Colo.
Trade names or commercial brands are mentioned solely
for information. No endorsement by the U.S. Department of
Agriculture is implied.
Sokal, Robert E., and F. James Rohlf. 1969. Biometry. W.
H. Freeman and Company, San Francisco, Calif. p. 518 ff.
The Author
FRANCIS M. FUJIOKA is a research meteorologist assigned to the
Hawaiian forest ecosystems work unit with headquarters in Honolulu,
Hawaii. He earned a B.S. degree (1968) and an M.S. degree (1972) in
meteorology a t the University of Hawaii. He joined the Forest Service and
the PSW staff in 1972.
U.S. Forest Service research in Hawaii
is conducted in cooperation with
Division of Forestry
Hawaii Department of Land and Natural Resources